zephyr-7b-sft-safeDPO / eval_results.json
AmberYifan's picture
End of training
c7d92f2 verified
raw
history blame contribute delete
571 Bytes
{
"epoch": 1.0,
"eval_logits/chosen": -1.7892440557479858,
"eval_logits/rejected": -1.7786749601364136,
"eval_logps/chosen": -389.2891845703125,
"eval_logps/rejected": -469.7640075683594,
"eval_loss": 0.5426768064498901,
"eval_rewards/accuracies": 0.719748854637146,
"eval_rewards/chosen": -2.265324115753174,
"eval_rewards/margins": 0.7290500402450562,
"eval_rewards/rejected": -2.9943742752075195,
"eval_runtime": 524.2847,
"eval_samples": 7000,
"eval_samples_per_second": 13.352,
"eval_steps_per_second": 0.418
}